Dynamic Authentication of Smartphone Users Based on Touchscreen Gestures

This paper proposes a dynamic authentication of smartphone users based on their gestures on touchscreen. The user authentication consists of three stages: collecting touch gestures' data, extracting features, and classification. Tapping, scrolling, dragging and zooming gestures' data are acquired using a developed android application. Then, features from these gestures are extracted. Finally, three different classifiers, medians vector proximity (MVP), k-nearest neighbor (k-NN) and random forest (RF), are applied to the extracted features. The performance of these classifiers are investigated and compared considering a single-touch gesture and all possible combinations of the extracted touch gestures on smartphone. The experimental results show that the MVP classifier brings the best results when using single gestures. When two gestures are combined together, the k-NN gives the best results. The k-NN classifier reaches an equal error rate of 0% using only three gestures. The RF is not ideal to be used on smartphone users' authentication as it gives the worst results.